general_config_for_your_dataset.txt

#stage0:
python main.py data/yourdataset --workspace yours/ -O --bound 1 --scale 1.0 --dt_gamma 0 --stage 0 --lambda_tv 1e-8 --iters 50000 --background random

#change the {workspace}/mesh_stage0/mesh_0.ply, to your mesh, and rename it: mesh_0.ply
#please make sure your mesh and the mesh_0.ply are in the same coordinate system, you can check them in Blender, Maya or such.

#stage1:
python main.py data/yourdataset --workspace yours/ -O --bound 1 --scale 1.0 --dt_gamma 0 --stage 1 --use_brdf --use_restir --lambda_kd 0.02 --lambda_ks 0.0001 --lambda_normal 0.0003 --lambda_edgelen 0.1 --lambda_nrm 0.0007 --lambda_rgb_brdf 0.05


# nvs eval using 512 spp
# note that validation_brdf will be overwritten every evaluation, so you can change current validation_brdf's name before new evaluation
# note that --roughness_min and --me_max should be the same as when trainning
python main.py data/yourdataset --workspace yours/ -O --bound 1 --scale 1.0 --dt_gamma 0 --stage 1 --use_brdf --use_restir --test --test_no_mesh --spp 512


# sometimes this config won't work with metal or specular scene, you can try with --me_max 0.2 or --me_max 0.4
# remember that --roughness_min and --me_max should be the same at evaluation as when trainning